Shows the genetic changes that diagnostic panels routinely miss — duplications, deletions, rearrangements, and rare mosaic variants present in just a few percent of cells. Built for long-read sequencing data; each variant stays visible individually rather than collapsed into a single line of a results table.
Some genes vary wildly in copy number — not in patients, in healthy people. And often within a single person. Standard sequencing routinely undercounts every region in this list.
Higher copies in starch-eating populations — recent human evolution caught in the act.
Part of the complement cascade. Reduced copy number is an established autoimmune-disease risk.
Antimicrobial peptides defending mucosal surfaces. Different people carry very different defensive arsenals.
Killer-cell receptors come in entirely different gene combinations between people. Decisive in transplantation and viral defense.
Plasma Lp(a) levels are determined by the number of internal kringle repeats — a direct cardiovascular-risk signal.
Common but previously unrecognized: IGHG locus duplications and triplications create hybrid IGHG1→IGHG4 fusion genes. Mothers carrying them transfer less IgG to the fetus — measurable as lower antibody levels in newborns and higher respiratory-infection risk in infancy. Reported in Belios et al., medRxiv 2025.
Are you confident standard short-read sequencing is seeing all of this? Tumor subclones, somatic mosaicism, immune-cell rearrangements — they all break the comfortable fiction of "one genome per person." VariantPaths is built for the regions where conventional tools quietly fail.
Four concepts, in case the field isn't your daily bread.
Not every genetic change is a single-letter typo. Sometimes whole stretches of DNA are duplicated, deleted, or rearranged. These structural variants drive many rare diseases, cancers, and developmental disorders — and conventional short-read sequencing (whether targeted panels or whole genomes) often misses them.
Older methods chop DNA into tiny fragments and reassemble them after the fact. Long-read platforms (PacBio, Oxford Nanopore) read very long stretches in one piece. The payoff: complex rearrangements that used to be invisible become visible.
Most tools collapse a complex region down to a single summary line in a results table. Important alternative versions vanish. VariantPaths keeps every variant as its own visible object, stacked on screen — so you can actually look at what is there.
VariantPaths is the companion viewer for BRANCH — a long-read assembler designed specifically to find low-frequency copy-number variants. BRANCH does the discovery; VariantPaths shows what is there. Together they make rare structural variants visible end-to-end.
Standard genome viewers reduce complex regions to a single row in a table. VariantPaths reads the bubble graphs that BRANCH produces — keeping every variant, every alternative path, every supporting read as its own visible object on screen.
Stacked rows per sample, like reads in IGV. See which variants are shared, which are unique, which appear in your control.
Below 5 bp per pixel the actual A/T/C/G letters appear inline. Inspect breakpoints, confirm exact insertion sequences, copy regions for primer design.
Right-click → MSA. Aligns every alternative version at that locus side by side. Distinguishes a real variant from a paralog mismatch.
Filter by frequency, length, recurrence. The 2–3 % mosaic range — tumor subclones, somatic mosaicism, early clonal events — is finally visible.
Every variant is annotated against dbVar, NCBI's public structural-variant database. "Already documented" or "never reported" is visible at a glance — no separate spreadsheet.
Single executable, no installer, no Python environment. Patient data never leaves the network — relevant for any setting with strict data-handling rules.
VariantPaths is in active development alongside the BRANCH assembler. Pre-built binaries aren't published yet — clone the repository and build from source.
git clone https://github.com/schlein-lab/variantpaths.git
cd variantpaths && cargo build --release
./target/release/variantpaths samples/igh_3sample.vpf samples/igh_3sample.vpz
Long-read sequencing reveals duplications, deletions, complex rearrangements, mosaic variants, and paralogous regions where one linear reference is no longer enough. VariantPaths shows them as they actually are.